Title: Dynamic Allocation in Honey Bee and Internet Server Colonies
1Dynamic Allocation in Honey Bee and Internet
Server Colonies
- Sunil Nakrani, Computing Lab., University of
Oxford, England, UK - Craig Tovey, ISyE, Georgia Institute of
Technology, Atlanta, USA
2Natural Systems Research Education
- Honey bee colony foraging (Bartholdi, Seeley,
Tovey VandeVate, J. Th. Bio. 1993) food
storing to cue nectar intake (Seeley Tovey,
Animal Beh. 1993) - Dominance hierarchy formation (Chase, Tovey, et
al., Proc. Nat. Acad. Sci 2002, Behaviour 2003)
natural selection mechanism - Biomimetic heuristic for allocating resources in
a web-hosting facility (Nakrani Tovey, Proc.
MASI II, 2003) - Time lags and overdiscounting of environmental
costs, hedging value of environmental
investments replacement policies under
technological change (Regnier, Sharp Tovey, IE
Trans.) - Assessing systems (Tovey, Ausenda) adjusting GDP
for natural systems deterioration - Sustainability intro in sophomore course (2030)
topics course on root causes of env. problems and
sustainability (4833) stat and design
sustainability projects
OR -gt BIO
BIO -gt OR
OR -gt ENV
3Introduction
- Web-Hosting Facility
- Rationale
- Benefits
- Server Allocation Problem allocate servers
amongst web-apps to maximize revenue - Honey Bee Colony allocate foragers amongst
flower patches to maximize nectar intake
4Introduction
- Approach Honey Bee Heuristics-waggle dance
- Map web-apps to flower patches, servers to bees
- Solution Mapping dance floor--gt advert board
- Algorithms and Simulation Model
- Results
- Conclusions, biological insight
- Future work
5Web-Hosting Facility
Internet
Hosting Center
Users
Web-App
6Web-Hosting Model
- Benefits
- Economy of scale Resource sharing means increase
in utilization and better availability - Web-App shielded from over-provisioning
7Web-Hosting Optimisation
- Web-App pay-per-use Service Level Agreement
(SLA) - Hosting Center Allocate servers among Web-Apps
to maximize revenue (s.t. changeover downtime) - Users Unpredictable and highly variable request
pattern
8Web-Hosting Optimisation
- Server Allocation Problem Allocate servers among
web-Apps to maximize revenue
9Server Allocation Problem
- Current Techniques Threshold and Ad-hoc Rule
based, Continuous tracking of load metrics by
large operations staff, Manual management - Static provisioning altered approx. once a month
- Current Literature Jayram et. al. (2001), Chase
et. al. (2001) - Commercial Domain Proprietary methods
10Honey Bee Colony
- Approx. 20-50 thousand bees in a colony
- One queen
- Few drones
- Rest workers
11Honey Bee Colony
- Typically requires 60 lb of honey per year to
survive - 25 of workers engaged in food collection
(nectar, pollen) - Exploit food sources (flower patches) from
surrounding countryside
12Honey Bee Colony
- Flower Patches
- Availability varies daily and seasonally
- Quality depends on exploitation, flower type,
micro-climate etc.. - Round trip time (nectar collection time)
- Colony Exploit flower patches efficiently to
satisfy nectar requirement
13Forager Allocation Problem
- Forager Allocation Problem Allocate forager bees
among flower patches to Maximize nectar intake
14Problem Mapping
- Server Allocation Problem
- Single Server
- Web-Apps User
- Group of servers (cluster) serving users at one
web-app
- Forager Allocation Problem
- Forager Bee
- Flower Patches
- Group of foragers collecting nectar at a specific
flower patch
15Problem Mapping
- Server Allocation Problem
- Request service time depends on Web-App
- Find a user to serve
- Forager Allocation Problem
- Travel Time depends on Flower Patch
- Nectar collection time at the patch
16Problem Mapping
- Server Allocation Problem
- Value-Per-Request-Served
- Varying rates of user request arrivals and
balking behaviors
- Forager Allocation Problem
- Nectar quality (sugar content)
- Varying flower patch density, quality, and
replenishment rate
17Problem Mapping
- Server Allocation Problem
- Server Migration Time (purge current Web-App and
load new Web-App)
- Forager Allocation Problem
- Time to learn the location of the flower patch
and successful discovery (Seeley, T.D.)
18Forager Allocation Mechanism
- Active foragers return to the hive with nectar
and profitability rating of the visited flower
patch - Interact with food-storer bees to offload nectar
(waiting time provides feedback on nectar flow
into the hive)
19Forager Allocation Mechanism
- Feedback sets threshold for enlisting signal
(Waggle Dance) - Profitability signal threshold Waggle dance
duration
20Forager Allocation Mechanism
- Waggle dance performed just inside the hive
entrance (Dance floor) - foragers follow dance to learn flower patch
location - Suboptimal allocation in static sense
21fi(xi) return from xi bees at patch i Max
åi fi(xi) s.t. xi 0 åi xi N
- OPTIMUM
- fi0(xi) l 8 i2 A
- xi 0 8 i Ï A
- equalize marginal return at active patches
- BEE HEURISTIC
- fi(xi)/xi m 8 i2 A
- xi 0 8 i Ï A
- equalize average return at active patches
22Properties of Heuristic Solution(from BSTV 93)
- Usually not optimal
- Factor-2 approximation even under very weak
conditions - Convergence proved by potential function argument
- Validated experimentally in a honey bee colony
23Solution Mapping
- Server Allocation
- Advert
- Advert Board
- Advert Duration
- Reading an Advert
- Forager Allocation
- Waggle Dance
- Dance Floor
- Dance Duration
- Following Waggle Dance
24Simulation Model Honey Bee
Web-App A
Post/Read Adverts
Users A
Web-App ID Duration Time
Advert Board
Repurpose
Migrate
Web-App ID Duration Time
Post/Read Adverts
Users B
Web-App B
25Simulation Model Greedy
Web-App A
Users A
New Policy
Compute optimal policy for next interval based
on present queue status, present allocation,
and user arrival from last interval
Repurpose
Migrate
Users B
New Policy
Web-App B
26Simulation Model Greedy
- St state of world at start period t
(customers,servers) - At arrivals (times, types) in period t
- P(p, S, A) profit using p from state S with
arrivals A - f(p,S,A) next state of world using p
- from S with arrivals A
- ptG arg maxp P(p, St, At-1)
- St1 f(ptG, St, At)
27Simulation Model Others
Web-App A
Users A
New Policy
Offline Omniscient Computation
Repurpose
Migrate
Users B
New Policy
Web-App B
28Simulation Model Omniscient Optimum
- Sstate, Aarrival, P( )profit, f( )next state
- A1,L, An known
- vn1(Sn1) 0 (no salvage value)
- vt (St) maxpP(p,St,At)
vt1(f(p,St,At)) - ptOpt(St) arg maxp P(p,St,At)
vt1(f(p,St,At))
29Omniscient Optimum Computation
- Parallel implementation runs in 24 hours
- Discretized space of possible states
- Inner loop function that we maximize is
theoretically concave - but not concave numerically
30(No Transcript)
31Simulation Model Optimal-Static
- Sstate, Aarrival, P( )profit, f( )next state
- A1,L, An known
- s.t. St1 f(p, St, At)
32Test Case Synthetic User Load
33Test Case Real Internet Trace
34Result Synthetic User Load
35Result Internet Service Trace Load
36Adaptability to Synthetic Variable Load
37Synthetic Load Low Variability
38Conclusions
- Bee heuristic works well, effective in highly
dynamic environment - Competitive against standard heuristics
- Bee heuristic Not tuned, Common sense scaling
parameters used
39Conclusions
- Trade-off static optimality for responsiveness
- Static optimization requires equalization of
derivatives (marginal rate bee) - Bee heuristic has no marginal bee but,
instead, has ability to migrate several bees at
the same time and avoids problem of measuring f
under variability
40Conclusions
Patch II
Patch I
900
500
Nectar intake increases if
899
501
41Future Work
- Test to see if we were lucky or robust
- Scale up to more patches/web-apps
- Make autonomic --more feedback loops
- Power imitate indolent bees?
- Convergence rates
- Compare with IBMs online network algorithm
42Some other interesting stuff
- Dominance hierarchies first experimental
validation of a self-organizing social structure
in animals (Chase, Tovey, Martin Manfredonia
02) - Time lags of environmental costs mean 10 years
vs. mean 5 years for other types. (Regnier
Tovey) - Opportunities for Sr. Design sustainability
projects
43Some Big OR Questions in Natural Systems
- Individual versus group selection classic
argument against latter is essentially an OR
proof, but why do forests thrive? - Discounting and EPV, intergenerational equity
and intraperiod utility. Relationship to future
growth? Intraperiod utility and discounting is
almost equivalent to linear utility, Sobel 2000